Disease-Gene: Self-referential cases

[joeflack4]

Disease-Gene: Self-referential somatic cancers
Self-referential “phenotype in the gene position + Phenotype field without a MIM” + "morbidmap.txt entry not in Phenotype-Gene Relationships table" case - add to review.tsv

All current cases: Google Sheet

[twhetzel]

For the code, see comment inline about hard-coding OMIM identifiers

[twhetzel]

Overall looks good! There are some comments/descriptions that will be more clear to our future selves if expanded a bit further and have some examples. I acknowledge these comments are a bit nit picky, but if the docs are being updated they should be clear without a lot of prior knowledge of all the deep review and understanding of the data files that went into identifying these cases.

[twhetzel]

@joeflack4 can you add a section that includes what a "valid disease-gene (D2G) relationship" is, ie the rules and that 'valid' means it is a gene association that is represented with RO:0004003 'has material basis in germline mutation in'?

[joeflack4]

There's a lot of commentary scattered throughout the codebase, but not really anything in the README / front page that explains this stuff. So it will entail a new section; I'm not sure what to call it. Maybe an "additional information" or "topics" or "explanation" section with subsection "Disease-Gene associations".

[twhetzel]

A new section in the README is fine

[joeflack4]

@twhetzel I updated the README. I'm not sure if I'm done yet; check it out.

[twhetzel]

I'd prefer to review when you have added everything. Is there some specific feedback you are looking for about the README? The additions I requested should only take ~20 min to add.

[joeflack4]

I'll finish it up and then tag you again. I took is taking a little more time than that. You've been wanting a lot of related documentation for a while, so I compiled it all in a way I felt you and the team would like, phrased in a way that is as clear / least confusing as I can make it (harder to do in this case than other things I typically document).

[joeflack4]

Just re-checked and these README.md updates are done.

• Addition of new section: Under the hood: Design decisions, etc.
  • Has only one sub-section: Gene-Disease Pipeline
    • Include's Trish's request: can you add a section that includes what a "valid disease-gene (D2G) relationship"
    • Additional content to support that disease-defining explanation, and non-disease defining stuff (copy/pasted from code notes and reformatted)
      • Example input/output
      • Inverse predicate for disease-defining (RO:0004003 and RO:0004013)
      • Other non-disease-defining gene-disease association predicates.

[twhetzel]

say how the lines are related

[joeflack4]

Instead I changed "and are related" to "and are part of a pattern".

I think that is more accurate. The next bullets, describing exactly the pattern of "row 1" and "row 2", along with the link to the spreadsheet with examples, should be sufficiently illustrative I think.

[twhetzel]

include an example of where the phenotype MIM is in these kinds of rows so someone can understand the case without having to read all the code

[joeflack4]

There is already the link to the google sheet. I think that's probably enough.

However, the formatting of the spreadsheet is not 100% the same as morbidmap.txt, and it can be hard to interrogate that file sometimes.

So I added this table as an example:

Phenotype	Gene/Locus And Other Related Symbols	MIM Number	Cyto Location
Small cell cancer of the lung, somatic, 182280 (3)	RB1	614041	13q14.2
Small-cell cancer of lung (2)	SCLC1	182280	3p23-p21

I think this will be sufficient to show how / where the Phenotype appears.

[twhetzel]

include an example data line so it's clear what is meant as the 'phenotype in the first row appears as a gene', which i think you mean the phenotype mim is in the position where a gene mim should be found

[joeflack4]

Same example just added see comment above should suffice.